NRPE race-condition caused by "chown" ran at update_nrpe_config and nrpe charm

[Pedro Guimarães]

Hi,

Running latest stable charms.
Some of nrpe checks on kubernetes-worker fail with:

Unable to run kubectl and parse output

That is caused whenever /var/lib/nagios is owned by root instead of nagios user.
That is a clear race condition between nrpe and worker charms.

On worker code, I can see:
https://github.com/charmed-kubernetes/charm-kubernetes-worker/blob/8c3c6061bec0e90ce1eaea24d570d5e7e19e8a78/reactive/kubernetes_worker.py#L1111

Where worker sets owner of the folder as nagios.
On nrpe charm, however, I see the folder is set for root:

2020-11-27 12:07:56 INFO juju-log Making dir /usr/local/lib/nagios/plugins root:root 755
2020-11-27 12:07:56 INFO juju-log Making dir /var/lib/nagios/export/ root:root 755

Therefore, some units may end up with root owner as well. That will depend on which hook is ran before.

This is also confirmed on stat command:
https://pastebin.ubuntu.com/p/RD7scS4G2Z/

Where failing units have root as owner.

[Pedro Guimarães (pguimaraes)]

Also adding NRPE charm, since it should be setting proper permissions on /var/lib/nagios as part of install hook.

[Chris Johnston (cjohnston)]

@Pedro, can you confirm which revision of the charms you are using? This should have been fixed in 1.19+ck1 as https://bugs.launchpad.net/charm-kubernetes-worker/+bug/1866382

[Eric Chen (eric-chen)]

Close this issue because of no update of this issue over 1 year.
Please consider using the new Canonical Observability Stack instead.
(https://charmhub.io/topics/canonical-observability-stack)